Date: 2006.7.28 (Fri) 16:30-17:45
Place: Building No.4, Research Center for Advanced Science and Technology
Speaker: Masatoshi Funabashi
Title: Modeling and Analysis of Birdsong Learning
Keywords: chaotic wandering, complex systems neuroscience, Tinbergen's four questions, finite-state syntax, Elman network

Affiliation: Institute of Industrial Science, University of Tokyo
Advisor: Kazuyuki Aihara
Disciplines: Mathematical Enginnering, Neuroscience, Ethology

Bibliography: Masatoshi Funabashi, Modeling and Analysis of Birdsong Learning, Human Science Integration Seminar Abstracts, No. 19, pp. 1, 2006.
(Please use this bibliography when you cite this abstract.)

Abstract:
Among passerines, Bengalese finches are known to sing extremely complex courtship songs, with 3 hierarchical structures: the element, the chunk, and the syntax. To elucidate the complex mechanism of the song of Bengalese finches, Okanoya and his colleagues have launched numerous studies based on Tinbergen's four questions for ethologists.
In chapter 2, to offer a more dynamic view on the development of birdsong learning, we first construct a model using the Elman network with chaotic neurons, that successfully learned the supervisor signal defined by a simple finite-state syntax.
Second, we focus on the process of individual-specific increases in the complexity of song syntax. We propose a new learning algorithm to produce the intrinsic diversification of song syntax without a supervisor based on the itinerant dynamics of chaotic neural networks and the Hebbian learning rule. The emergence of a novel syntax modifying the acquired syntax is demonstrated.
In chapter 3, to elucidate a more precise dynamics of chaotic Elman network (CEN), we investigate and compare the dynamics of chaotic neural network (CNN) to that of CEN. In a - kr plane of CEN Hidden layer, there exists three divisions of dynamics that are qualitatively different: Area I is the faithful retrieval of programmed patterns, while area II shows chaotic wandering process or periodic dynamics among stored patterns and spurious memories. Dynamics in area III is only periodic among spurious memories. In area II, there exists certain region where the chunks of syntax are relatively well conserved. By evaluating the moment Lyapunov exponents on an invariant subspace, we acquire a qualitative diagram to understand the dynamics of CEN.
The significance of this model is discussed.

References:
[1] Obana I., Aihara K., Fukui Y., and Hoshino H.: 1998, “Design of Chaotic Responses in Neuron model and its Application to Modeling of Artificial Living Things of Moving to Light on a Plane.” The Journal of the Institute of Electronics, Information and Communication Engineers, MBE 88-7, pp. 43-50.
[2] Aihara K., Takabe T. and Toyoda T.: 1990, “Chaotic neural networks.” Phys. Lett. A, 144, pp. 333-340.
[3] Chen L. and Aihara K.: 1995, “Chaotic Simulated Annealing by a Neural Network Model with Transient Chaos.” Neural Networks, Vol.8, No.6, pp.915-930.
[4] Elman J. L.: 1990, “Finding Structure in Time.” Cognitive Science, 14, pp. 179-211.
[5] Fujii H., Aihara K. and Tsuda I.: 2004, “Functional Relevance of ‘Excitatory’ GABA Actions in Cortical Interneurons: A Dynamical Systems Approach.” Journal of Integrative Neuroscience, Vol.3, No.2, pp. 183-205.
[6] Kawamura T. and Okanoya K.: 2001, “The variable N-gram as a model of the brain representation for the sequential behavior.” International Congress of Neuroethology, p398.
[7] Kitajima H.: 2003, “Itinerant memory dynamics and global bifurcations in chaotic neural networks.” CHAOS, Vol. 13, No. 3, pp. 1122-1132.
[8] Komuro M. and Aihara K.: 2001, “Hierarchical Structure among Invariant Subspaces of Chaotic Neural Networks.” Japan Journal of Industrial and Applied Mathematics, Vol.18, No.2, pp.335-357.
[9] Kuroiwa J., Matsunami N., Nara S., and Aihara K.: 2004 “Sensitive Response of a Chaotic Wandering State to Memory Fragment Inputs in a Chaotic Neural Network Model.” International Journal of Bifurcation and Chaos, Vol.14, No. 4, pp.1413-1421.
[10] Okanoya K. and Yoneda T.: 1995, “Phonetic Development of Avian Species -Analysis by an Analogy with Neural Networks-.”(in Japanese) Comparative Physiology and Biochemistry (in Japanese), Vol. 12, No. 1, pp. 3-13.
[11] Okanoya K.: 2002, The Transition to Language, Oxford University Press, Oxford.
[12] Okanoya K.: 2003, From Birdsong to Human Language. (in Japanese), Iwanami, Tokyo.
[13] Sasahara K.: 2005, “Evolution of Complexity and Diversity in Simulated Birdsong Grammer.” PhD. Thesis at the Department of General Systems Studies, Graduate School of Arts and Sciences, The University of Tokyo.
[14] Servan-Schreiber D., Cleeremans A., and McClelland J.L.: 1988, “Encoding sequential structure in simple recurrent networks.” Technical Report CMU-CS-88-183, Carnegie Mellon University, Pittsburgh, PA.
[15] Skarda C. A. and Freeman W. J.: 1987, “How brains make chaos in order to make sense of the world.” Behavioral and Brain Sciences, 10, pp. 161-195.
[16] Tinbergen, N.: 1963, “On aims and methods of ethology.” Zeitschrift fur Tierpsychologie, 20, pp. 410-433.
[17] Tsuda I.: 1992, “Dynamic Link of Memory -Chaotic Memory Map in Nonequilibrium Neural Networks.” Neural Networks, 5, pp. 313-326.